From 83eab939a4f85b8571cee5c3907ac3de0f642f06 Mon Sep 17 00:00:00 2001
From: Richard Osborne <richard@xmos.com>
Date: Thu, 27 Feb 2014 13:20:11 +0000
Subject: [PATCH] [XCore] Add dag combines for instructions that ignore some
 input bits.

These instructions ignore the high bits of one of their input operands -
try and use this to simplify the code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202394 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/XCore/XCoreISelLowering.cpp  | 42 +++++++++++++++++++++++++
 test/CodeGen/XCore/resources_combine.ll | 41 ++++++++++++++++++++++++
 2 files changed, 83 insertions(+)

diff --git a/lib/Target/XCore/XCoreISelLowering.cpp b/lib/Target/XCore/XCoreISelLowering.cpp
index 930a4d19ba0..450f0771d90 100644
--- a/lib/Target/XCore/XCoreISelLowering.cpp
+++ b/lib/Target/XCore/XCoreISelLowering.cpp
@@ -180,6 +180,8 @@ XCoreTargetLowering::XCoreTargetLowering(XCoreTargetMachine &XTM)
   // We have target-specific dag combine patterns for the following nodes:
   setTargetDAGCombine(ISD::STORE);
   setTargetDAGCombine(ISD::ADD);
+  setTargetDAGCombine(ISD::INTRINSIC_VOID);
+  setTargetDAGCombine(ISD::INTRINSIC_W_CHAIN);
 
   setMinFunctionAlignment(1);
   setPrefFunctionAlignment(2);
@@ -1566,6 +1568,46 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N,
   SDLoc dl(N);
   switch (N->getOpcode()) {
   default: break;
+  case ISD::INTRINSIC_VOID:
+    switch (cast<ConstantSDNode>(N->getOperand(1))->getZExtValue()) {
+    case Intrinsic::xcore_outt:
+    case Intrinsic::xcore_outct:
+    case Intrinsic::xcore_chkct: {
+      SDValue OutVal = N->getOperand(3);
+      // These instructions ignore the high bits.
+      if (OutVal.hasOneUse()) {
+        unsigned BitWidth = OutVal.getValueSizeInBits();
+        APInt DemandedMask = APInt::getLowBitsSet(BitWidth, 8);
+        APInt KnownZero, KnownOne;
+        TargetLowering::TargetLoweringOpt TLO(DAG, !DCI.isBeforeLegalize(),
+                                              !DCI.isBeforeLegalizeOps());
+        const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+        if (TLO.ShrinkDemandedConstant(OutVal, DemandedMask) ||
+            TLI.SimplifyDemandedBits(OutVal, DemandedMask, KnownZero, KnownOne,
+                                     TLO))
+          DCI.CommitTargetLoweringOpt(TLO);
+      }
+      break;
+    }
+    case Intrinsic::xcore_setpt: {
+      SDValue Time = N->getOperand(3);
+      // This instruction ignores the high bits.
+      if (Time.hasOneUse()) {
+        unsigned BitWidth = Time.getValueSizeInBits();
+        APInt DemandedMask = APInt::getLowBitsSet(BitWidth, 16);
+        APInt KnownZero, KnownOne;
+        TargetLowering::TargetLoweringOpt TLO(DAG, !DCI.isBeforeLegalize(),
+                                              !DCI.isBeforeLegalizeOps());
+        const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+        if (TLO.ShrinkDemandedConstant(Time, DemandedMask) ||
+            TLI.SimplifyDemandedBits(Time, DemandedMask, KnownZero, KnownOne,
+                                     TLO))
+          DCI.CommitTargetLoweringOpt(TLO);
+      }
+      break;
+    }
+    }
+    break;
   case XCoreISD::LADD: {
     SDValue N0 = N->getOperand(0);
     SDValue N1 = N->getOperand(1);
diff --git a/test/CodeGen/XCore/resources_combine.ll b/test/CodeGen/XCore/resources_combine.ll
index 50ac30e905b..20c184a53b9 100644
--- a/test/CodeGen/XCore/resources_combine.ll
+++ b/test/CodeGen/XCore/resources_combine.ll
@@ -5,6 +5,10 @@ declare i32 @llvm.xcore.inct.p1i8(i8 addrspace(1)* %r)
 declare i32 @llvm.xcore.testct.p1i8(i8 addrspace(1)* %r)
 declare i32 @llvm.xcore.testwct.p1i8(i8 addrspace(1)* %r)
 declare i32 @llvm.xcore.getts.p1i8(i8 addrspace(1)* %r)
+declare void @llvm.xcore.outt.p1i8(i8 addrspace(1)* %r, i32 %value)
+declare void @llvm.xcore.outct.p1i8(i8 addrspace(1)* %r, i32 %value)
+declare void @llvm.xcore.chkct.p1i8(i8 addrspace(1)* %r, i32 %value)
+declare void @llvm.xcore.setpt.p1i8(i8 addrspace(1)* %r, i32 %value)
 
 define i32 @int(i8 addrspace(1)* %r) nounwind {
 ; CHECK-LABEL: int:
@@ -50,3 +54,40 @@ define i32 @getts(i8 addrspace(1)* %r) nounwind {
 	%trunc = and i32 %result, 65535
 	ret i32 %result
 }
+
+define void @outt(i8 addrspace(1)* %r, i32 %value) nounwind {
+; CHECK-LABEL: outt:
+; CHECK-NOT: zext
+; CHECK: outt res[r0], r1
+; CHECK-NEXT: retsp 0
+	%trunc = and i32 %value, 255
+	call void @llvm.xcore.outt.p1i8(i8 addrspace(1)* %r, i32 %trunc)
+	ret void
+}
+
+define void @outct(i8 addrspace(1)* %r, i32 %value) nounwind {
+; CHECK-LABEL: outct:
+; CHECK-NOT: zext
+; CHECK: outct res[r0], r1
+	%trunc = and i32 %value, 255
+	call void @llvm.xcore.outct.p1i8(i8 addrspace(1)* %r, i32 %trunc)
+	ret void
+}
+
+define void @chkct(i8 addrspace(1)* %r, i32 %value) nounwind {
+; CHECK-LABEL: chkct:
+; CHECK-NOT: zext
+; CHECK: chkct res[r0], r1
+	%trunc = and i32 %value, 255
+	call void @llvm.xcore.chkct.p1i8(i8 addrspace(1)* %r, i32 %trunc)
+	ret void
+}
+
+define void @setpt(i8 addrspace(1)* %r, i32 %value) nounwind {
+; CHECK-LABEL: setpt:
+; CHECK-NOT: zext
+; CHECK: setpt res[r0], r1
+	%trunc = and i32 %value, 65535
+	call void @llvm.xcore.setpt.p1i8(i8 addrspace(1)* %r, i32 %trunc)
+	ret void
+}
-- 
2.34.1